The Judgment That Never Reached the Question

I.The First Known Thing — the loss that binds, and the loss that does not

law · civil procedure, issue preclusion

A legal system that let every question be reopened would never finish anything, and one that let every loss settle everything would be unjust. The doctrine that sits between these, known in the older books as collateral estoppel and in the modern ones as issue preclusion, says when a point decided in one proceeding may be treated as closed in a later one. Its conditions are stated compactly in the Restatement (Second) of Judgments: the issue in the second proceeding must be the same issue; it must have been actually litigated; it must have been actually determined; and the determination must have been essential to the judgment. Miss any one and the earlier loss, however expensive and however public, binds nobody on that point.

each condition excludes a different kind of loss

The conditions are not a formality, and each rules out a familiar and superficially convincing case. A default judgment, or one entered on admission or consent, decides the case but litigates no issue, so nothing in it precludes. A dismissal for want of jurisdiction, for improper venue, or for failure to prosecute is a loss on the merits of nothing, and the claim may be brought again elsewhere. An issue raised and briefed but disposed of on another ground was not determined. And the requirement of essentiality does the subtlest work of all. A finding that the court did not need in order to reach its result — obiter, an aside, a second reason offered for completeness — carries no preclusive force, on the reasoning that a point which changed no outcome was not scrutinised with the care that finality demands, and, crucially, that the winner on that point had no occasion to appeal it. The same logic produces the rule for alternative determinations: where a judgment rests on two independent grounds, each sufficient on its own, the Restatement's position is that neither is preclusive, precisely because neither was necessary and the loser's appeal of one would have availed nothing.

extending a loss to strangers is the strong form

The doctrine grew more powerful when the old requirement of mutuality fell. Historically a party could invoke a prior judgment only against someone who had been bound by it; a stranger to the first case could take no benefit from it. In Blonder-Tongue Laboratories v. University of Illinois Foundation (1971) the Supreme Court abandoned that requirement for a particular and instructive situation: a patentee who has litigated the validity of a patent and lost may not relitigate validity against a fresh defendant who was nowhere near the first case. One loss, and the proposition is closed against all comers. But the Court did not make this automatic. It attached an inquiry into whether the patentee had a full and fair opportunity to litigate the first time, and named the kinds of circumstance — an unfavourable forum, a want of crucial evidence, an incompetently tried case — in which the earlier loss should not be given that reach. Parklane Hosiery Co. v. Shore (1979) extended non-mutual preclusion to offensive use while making it discretionary, and set out the factors a court must weigh: whether the loser could have joined the earlier action, whether the stakes there were so small that the loser had little incentive to defend vigorously, whether there exist prior inconsistent judgments, and whether procedural opportunities available now were unavailable then.

what the discipline makes of it

Read together, the doctrine encodes a proposition that is not obvious and that the law arrived at by slow attrition: the weight a defeat deserves to carry beyond its own case is not a function of how large, how costly, how well reported or how final that defeat was. It is a function of four structural features of the defeat — what question was actually in contest, whether that question was decided on evidence, whether the decision was load-bearing, and whether the loser had the means and the motive to fight it. A catastrophic and widely reported loss can be worth nothing as precedent on a point; a modest one can close the point permanently. The literature that discusses this is civil procedure, its concerns are docket economy, repose and the risk of inconsistent judgments, and it has no interest whatever in what happens to a scientific hypothesis after an experiment goes badly.

II.The Second Known Thing — the class that empties after one failure

drug discovery · portfolio direction, programme termination and herding

Industrial drug discovery is organised around targets and mechanisms, and sponsors watch each other closely because the information is scarce and the trials are dear. The consequence is a characteristic dynamic that everyone in the business can describe: a target class attracts many programmes at once, and then a single large failure empties it. Entry into the class collapses, internal programmes are shelved, the mechanism acquires a reputation, and reviewers, investors and committees treat the proposition as settled. Nothing formal happens. There is no register of closed hypotheses and no body that declares one. The abandonment is made of many separate decisions taken by people reading the same readout.

the type case

The clearest instance is inhibition of cholesteryl ester transfer protein. Torcetrapib was taken into a large cardiovascular outcomes trial and stopped in 2006 for excess mortality, reported in the New England Journal of Medicine the following year. The class emptied. Yet within two years it was shown that the compound raised blood pressure and circulating aldosterone by a mechanism independent of CETP inhibition — an off-target property of that molecule, demonstrated in cells and animals lacking the target altogether. On the question the field took the trial to have answered, namely whether inhibiting CETP benefits patients, the trial had determined nothing; it had determined that this compound harmed them, for a reason that has no bearing on the class. Two later inhibitors failed for their own reasons and a third produced a modest effect, and only two decades on has the class been revisited in earnest, with obicetrapib now producing substantial lowering of low-density lipoprotein in large trials and a literature that has begun to describe the episode as a renaissance. Anti-amyloid therapy in Alzheimer's disease supplies a second instance with a different shape: a run of failures in symptomatic patients was widely read as refuting the amyloid hypothesis, when a competing reading held that the question had not been reached, because the antibodies used had not removed enough plaque and had been given too late; the later approvals of agents that do remove plaque did not settle that argument but did show that the class had been closed before it was tested.

what is recorded, and what is not

The discipline is not naive about termination. It knows that programmes stop for reasons unrelated to the science — strategic reprioritisation, a merger, a financing window, an enrolment failure — and the trial registries and commercial pipeline databases carry a field for the reason a study stopped, usually a short free text later binned into efficacy, safety, business or other. There is a body of work on wrong kills, on assets terminated by one sponsor and later approved by another, and frameworks intended to make the decision more disciplined. What none of this does is grade a failure by what it was capable of settling. The binary that the databases record is why this programme stopped. The quantity that governs the behaviour of everyone else is what this stopping proved about the mechanism, and it is nowhere computed. A trial halted for an off-target toxicity and a trial halted because the mechanism was engaged and did no good are both recorded as terminations for safety or for efficacy, are both reported as the failure of a named drug against a named target, and are both read by the rest of the field as the same kind of news.

III.The Bridge

no procedure treatise mentions a target class; no portfolio review cites a preclusion case

The halls do not meet. One is a body of procedural doctrine whose literature is judgments, restatements and law review commentary, and whose purpose is the orderly closing of disputes. The other is pharmaceutical research management, whose literature is attrition statistics, pipeline databases and decision frameworks. Neither cites the other, no review spans them, and the shared vocabulary — evidence, judgment, finality — is ordinary usage on one side and technical on the other. Yet both describe the same situation: a costly, public, non-repeatable contest whose outcome is then taken by parties who were not in it to have settled a general proposition, when the contest may not have put that proposition in issue at all.

X: the reach of a defeat is set by what it decided, not by how large it was

Call it preclusive weight. Where a costly contest is held once and its outcome is then treated by parties who were not in it as closing a general proposition, the weight that outcome deserves is not a function of its magnitude, its cost, its prominence or its finality, but of four structural conditions, each of which can fail independently. First, identity: was the general proposition the thing actually in contest, or was the contest about a particular — this molecule, this dose, this population — from which the general proposition merely happened to be inferred. Second, determination: was the proposition decided on evidence bearing upon it, or disposed of on some other ground that made it unnecessary to reach. Third, necessity: was the proposition load-bearing for the outcome, or one of two independent sufficient grounds, in which case neither was pressed to destruction and neither would have been worth contesting even by the party it harmed. Fourth, adequacy: did the losing party have the resources, the stakes and the room to contest it fully, since a proposition lost by a party with no incentive to defend it is not a proposition that has been tested. To these the law adds a warning about the strong form. Extending one party's loss against all other parties — which is exactly what a field does when a single sponsor's trial closes a mechanism for everyone — is the non-mutual move, and the common law permitted it only after centuries of refusal, only with the adequacy inquiry attached, and, in its offensive form, only at a court's discretion. A discipline that performs it automatically, on every failure, with no inquiry at all, is applying the strongest available doctrine under the loosest available conditions. The consequence is a definite and uncomfortable prediction: abandonment should track the visibility of a failure while revival should track its preclusive weight, so that the classes most worth reopening are systematically those closed by the failures that decided least.

IV.The Question That Becomes Possible

neither discipline would ask this alone

The procedural lawyer has no reason to ask it because there is no judgment in the question and no court to bind. The portfolio committee has no reason to ask it because the field's vocabulary for a failed trial has one axis, which is how bad the result was, and no axis for what the result was competent to settle. Held together, they produce a question with a definite shape and, unusually, with data already in existence.

the preclusive weight hypothesis

For a landmark failure that empties a target class — a first large trial of a first-in-class mechanism that stops badly — can a preclusive weight be scored from the four conditions, each coded from the protocol and the primary publication alone: whether the mechanism rather than the molecule was in issue; whether target engagement was demonstrated at the exposures reached, so that the mechanism was actually put to the evidence; whether the stopping rested on a single ground or on independent sufficient ones; and whether the sponsor had the means to run a trial capable of deciding the question. And does that score predict what the field then does, and what the mechanism later turns out to be worth, better than the quantities the field already has — the size and phase of the trial, the prominence of the sponsor, and the registry's existing bin of efficacy, safety or business? The sharp form is a divergence claim: the collapse in new entry after a failure should be governed by the failure's visibility and be largely indifferent to its preclusive weight, while the later revival and eventual success of the mechanism should be governed by preclusive weight and be largely indifferent to visibility. If both are true, the gap between them is a standing and measurable misallocation, and it can be pointed at in advance.

why it would matter

Because it converts a judgement that is currently made by reputation into one made from four features of a document, available on the day the failure is published; because it identifies, among the mechanisms a field has written off, the subset written off by trials that decided nothing about them, which is a searchable list rather than a matter of taste; and because it supplies the missing question for a committee looking at a dead class, which is not whether the trial was convincing but what the trial was competent to settle. And if it fails — if revival tracks nothing but the money and the mood, or if the registry's existing three-way bin predicts it just as well — then the field's abandonment behaviour is either rational or simply financial, and the doctrine has nothing to add to it.

V.How It Could Be Proven — or Broken

a candidate must be breakable, or it is merely poetry

An intercognix candidate earns nothing until it names its own test. This one is retrospective in the first instance, which makes it cheap and also makes it fragile in a way that has to be handled explicitly, since the outcome is known to anyone who codes the inputs.

the honest status

Note the narrowness. The claim concerns class-level abandonment after a first pivotal failure of a shared and named mechanism, and nothing else. It says nothing about decisions on individual compounds, where the reasons are many and mostly local; it does not assert that any particular abandoned class is worth reviving, and it emphatically does not assert that the field abandons classes too readily in general, since most failures are highly preclusive and most dead targets are dead for good reason. The claim is narrower and stranger than that: that the field has no way of telling the two kinds of failure apart, and that a doctrine built for an entirely different purpose already contains the distinction. The analogy is also imperfect in a way worth stating plainly rather than burying. A judgment is an authoritative determination by an adjudicator on a closed record, while a trial result is evidence, and the proper response to evidence is to update rather than to close. Preclusion exists for repose and docket economy, not for truth, and importing it wholesale would be importing a logic of finality into a place where finality is the error being diagnosed. What transfers is not the doctrine's purpose but its taxonomy: four conditions under which a defeat does not decide what it appears to decide, arrived at by a discipline that has been sorting defeats for several centuries, offered to one that has been sorting them for fifty years and records only how bad they were. Until the scores have been coded blind and compared against the registry's own bin, this remains exactly what its stamp says: a candidate. A structural echo between two halls of the library, waiting for someone with the tools of both to walk the corridor.